The rapidly increasing demands for computational throughput, bandwidth, and memory capacity fueled by breakthroughs in machine learning pose substantial challenges for conventional electronic computing platforms. For digital scaling to keep pace with the accelerating growth of artificial intelligence (AI) models beyond the trajectory of Moores law, computational power has to double roughly every three months. Historically, advancing compute performance relied on spatial scaling to increase the transistor count on a given chip area and, more recently, the development of parallel and multi-core architectures. Exponential scaling on trajectories much steeper than what can be achieved by such conventional strategies, and in line with the demands of AI, can be achieved with computing platforms that process data using multiple, orthogonal dimensions available to photons. Here we elucidate pivotal developments in the realization of multidimensional computing platforms based on photonic systems. Moving to such architectures holds enormous promise for low-latency, high-bandwidth information processing at reduced energy consumption.

The combination of metasurfaces with chalcogenide phase-change materials is a highly promising route toward the development of multifunctional and reconfigurable nanophotonic devices. However, their transition into real-world devices is hindered by several technological challenges. This includes, amongst others, the lack of large area photonic architectures produced via scalable nanofabrication methods as required for free-space photonic applications, along with the ability to withstand the high temperatures needed for the phase-change process. In this work, we present a scalable nanofabrication strategy for the production of reconfigurable metasurfaces based on high-throughput, large-area nanoimprint lithography that is fully compatible with chalcogenide phase-change materials processing. Our approach involves the direct imprinting of high-melting-point, thermally stable TiO2 nanoparticle pastes, followed by the deposition of an Sb2Se3 thin film as the phase-change material active layer. The patterned TiO2 film enables the creation of thermally robust metasurfaces, overcoming the limitations of conventional polymer-based nanoimprinting techniques. The versatility of our approach is showcased by producing phase-change devices with two distinct functionalities: (i) metasurfaces with tunable spectral band switching and amplitude modulation capabilities across the near- to mid-infrared, and (ii) reconfigurable chiral metasurfaces, whose chiroptical activity can be switched between the visible and the near-infrared. Experimental results show excellent agreement with numerical simulations and reveal high uniformity across large areas. This work provides a universal, thermally robust and scalable platform for the production of reconfigurable metasurfaces based on phase-change materials, paving the way to low-cost, photonic devices with reconfigurable optical responses that could be extended far beyond the applications demonstrated here.

In this work we demonstrate two phase-change material based actively controllable metasurfaces for optical mode conversion and control of optical orbital angular momentum. Such control finds a variety of important applications across technologically relevant fields including information processing, communication and quantum optics, and the need for versatile control devices in a compact and lower power form-factor is ever increasing [1]. The metasurface architecture used here exploits high contrast phase-change materials (PCMs) as the active component, allowing for very different metasurface functionalities upon switching of the PCM meta-atoms and offering the advantages of non-volatile, low-power and fast switching [2].

Optical analogue computing is attracting a lot of research attention due to some inherent advantages over digital electronic computing, advantages that include spatial parallelism, low-loss transmission and ultra-high bandwidth [1]. Both free-space and integrated photonic systems have been explored for uses such as optical neural networks [2] and parallelised matrix-vector multiplication [3]. The work presented here utilises a free-space configuration for reconfigurable optical image processing, in particular implementing dual-function reconfigurable low and high-pass filtering, with applications in edge-detection and blurring.

Probabilistic computing excels in approximating combinatorial problems and modeling uncertainty. However, using conventional deterministic hardware for probabilistic models is challenging: (pseudo) random number generation introduces computational overhead and additional data shuffling. Therefore, there is a pressing need for different probabilistic computing architectures that achieve low latencies with reasonable energy consumption. Physical computing offers a promising solution, as these systems do not rely on an abstract deterministic representation of data but directly encode the information in physical quantities, enabling inherent probabilistic architectures utilizing entropy sources. Photonic computing is a prominent variant of physical computing due to the large available bandwidth, several orthogonal degrees of freedom for data encoding and optimal properties for in-memory computing and parallel data transfer. Here, we highlight key developments in physical photonic computing and photonic random number generation. We further provide insights into the realization of probabilistic photonic processors and their impact on artificial intelligence systems and future challenges.

The rapidly increasing demands on computational throughput, bandwidth and memory capacity fuelled by breakthroughs in machine learning pose substantial challenges for conventional electronic computing platforms. Historically, advancing compute performance relied on miniaturization to increase the transistor count on a given chip area and, more recently, on the development of parallel and multicore architectures. Computing platforms that process data using multiple, orthogonal dimensions can achieve exponential scaling on trajectories much steeper than what is possible with conventional strategies. One promising analog platform is photonics, which makes use of the physics of light, such as sensitivity to material properties and ability to encode information across multiple degrees of freedom. With recent breakthroughs in integrated photonic hardware and control, large-scale photonic systems have become a practical and timely solution for data-intensive, real-time computational tasks. Here, we explain developments in the realization of multidimensional computing platforms based on photonic systems. Moving to such architectures holds promise for low-latency, high-bandwidth information processing at reduced energy consumption. Multidimensional photonic computing is a framework that combines classical and quantum approaches, leveraging the properties of light. This Perspective explores its potential to enable scalable, neuromorphic photonic quantum systems suited to data-intensive and complex computational tasks.

The next generation of smart imaging and vision systems will require compact and tunable optical computing hardware to perform high-speed and low-power image processing. These requirements are driving the development of computing metasurfaces to realize efficient front-end analog optical pre-processors, especially for edge-detection capability. Yet, there is still a lack of reconfigurable or programmable schemes, which may drastically enhance the impact of these devices at the system level. Here, we propose and experimentally demonstrate a reconfigurable flat optical image processor using low-loss phase-change nonlocal metasurfaces. The metasurface is configured to realize different transfer functions in spatial frequency space, when transitioning the phase-change material between its amorphous and crystalline phases. This enables edge detection and bright-field imaging modes on the same device. The metasurface is compatible with a large numerical aperture of 0.5, making it suitable for high resolution coherent optical imaging microscopy. The concept of phase-change reconfigurable nonlocal metasurfaces may enable emerging applications of artificial intelligence-assisted imaging and vision devices with switchable multitasking.

Integrated phase-change photonic devices, consisting of sub-micrometre phase-change cells (e.g., Ge2Sb2 Te5, or GST) on top of optical waveguides, offer a novel route to non-volatile optical memory and computing. In the pursuit of reduced switching energies and increased switching speeds, plasmonic enhancement is a promising way to improve device performance [1], [2]. Here, we explore further this concept of plasmonic enhancement, concentrating on device designs with relatively simple architectures that lend themselves to easy manufacture.

Active metamaterials are engineered structures that possess novel properties that can be changed after the point of manufacture. Their novel properties arise predominantly from their physical structure, as opposed to their chemical composition and can be changed through means such as direct energy addition into wave paths, or physically changing/morphing the structure in response to both a user or environmental input. Active metamaterials are currently of wide interest to the physics community and encompass a range of sub-domains in applied physics (e.g. photonic, microwave, acoustic, mechanical, etc.). They possess the potential to provide solutions that are more suitable to specific applications, or which allow novel properties to be produced which cannot be achieved with passive metamaterials, such as time-varying or gain enhancement effects. They have the potential to help solve some of the important current and future problems faced by the advancement of modern society, such as achieving net-zero, sustainability, healthcare and equality goals. Despite their huge potential, the added complexity of their design and operation, compared to passive metamaterials creates challenges to the advancement of the field, particularly beyond theoretical and lab-based experiments. This roadmap brings together experts in all types of active metamaterials and across a wide range of areas of applied physics. The objective is to provide an overview of the current state of the art and the associated current/future challenges, with the hope that the required advances identified create a roadmap for the future advancement and application of this field.

The development of novel, compact, and reconfigurable devices for optical analog computing would pave the way for the next generation of imaging systems free from high power consumption electronics and computationally demanding processing algorithms. Recently, nonlocal metasurfaces have emerged as a powerful platform to perform analog image processing operations with low energy consumption, at the speed of light, and without the need to physically access the Fourier space, thereby providing both high computational speeds and ease of integration. However, once such devices are designed and fabricated, their effect on optical beams is fixed, constraining their performance to a singular function. Here, we show how nonlocal metasurfaces made of novel low-loss chalcogenide phase-change materials, such as Sb 2 Se 3 , offer a degree of reconfigurability, enabling switching between certain imaging modes. Specifically, we show that switching between a two-dimensional edge-detection mode and a bright-field imaging mode, or between a two-dimensional edge-detection mode and a two-dimensional image blurring mode, is possible.

The potential for realizing fast, energy-efficient integrated photonic memory and computing devices developed from the nanoscale light-squeezing and electric-field enhancing capability of plasmonic resonant structures and the intrinsic tuneability of chalcogenide phase-change materials is explored. We concentrate on designs that should be readily manufacturable, comprising a plasmonic dimer-bar nanoantenna deposited on top of a phase-change cell, itself deposited on top of an integrated photonic waveguide. Device optical properties and switching behavior are determined by a combination of finite-element thermo-optic and bespoke phase-change computational models. The results show that suitably designed devices can achieve switching energies in the tens of pico-Joule range and switching speeds in the tens of nanosecond range, a very considerable improvement over conventional designs, and showing a good trade-off between the device performance and fabrication complexity.

Metasurfaces based on chalcogenide phase-change materials offer a highly promising route towards the realization of non-volatile reconfigurable metasurfaces. However, since their switching mechanism between amorphous and crystalline states is based on thermal stimuli, phase-change metasurfaces should be treated carefully when operating under high power laser sources, since optically induced heating could trigger unwanted state changes during their operation. In this work, therefore, we develop a thermodynamic model capable of tracking the crystallization, melting and reamorphization dynamics of phase-change optical metadevices, and so too their optical performance, when operating under (i.e., aiming to control) high power laser sources. Our model is used, by way of example, to ascertain the optical power-handling capabilties of two typical phase-change metasurface architectures, one for beam steering and one for active lensing.

Phase-change materials can deliver active metasurfaces, but it is not well-studied how these devices perform in high-power applications. We develop a model for optical performance changes induced by high-power lasers, applying it to two metasurface designs.

Flexible electronics which are easy to manufacture and integrate into everyday items require suitable memory technology that can function on flexible surfaces. Herein, the properties of Ge‐rich GeSbTe (GST) and Se‐substituted GeSbSeTe (GSST) phase‐change alloys are investigated for application as nonvolatile write‐once and rewritable memories in flexible electronics. These materials have a higher crystallization temperature than the archetypal composition of Ge 2 Sb 2 Te 5 and hence better data retention properties. Moreover, their high crystallization temperature provides for a particularly straightforward implementation of a write‐once memory configuration. Material properties of Ge‐rich GST and GSST are measured as a function of temperature using four‐point probe electrical testing, Raman spectroscopy, and X‐ray diffraction. Following this, the switching of flexible memory devices is investigated through both simulation and experiment. More specifically, crossbar memory devices fabricated using Ge‐rich GST are experimentally fabricated and tested, while the operation of GSST pore cell structures suitable for flexible memory applications is demonstrated through simulation.

Biological neural networks effortlessly tackle complex computational problems and excel at predicting outcomes from noisy, incomplete data, a task that poses significant challenges to traditional processors. Artificial neural networks (ANNs), inspired by these biological counterparts, have emerged as powerful tools for deciphering intricate data patterns and making predictions. However, conventional ANNs can be viewed as "point estimates" that do not capture the uncertainty of prediction, which is an inherently probabilistic process. In contrast, treating an ANN as a probabilistic model derived via Bayesian inference poses significant challenges for conventional deterministic computing architectures. Here, we use chaotic light in combination with incoherent photonic data processing to enable high-speed probabilistic computation and uncertainty quantification. Since both the chaotic light source and the photonic crossbar support multiple independent computational wavelength channels, we sample from the output distributions in parallel at a sampling rate of 70.4 GS/s, limited only by the electronic interface. We exploit the photonic probabilistic architecture to simultaneously perform image classification and uncertainty prediction via a Bayesian neural network. Our prototype demonstrates the seamless cointegration of a physical entropy source and a computational architecture that enables ultrafast probabilistic computation by parallel sampling.

The design, fabrication and characterisation of an amplitude-only, constant-phase spatial light modulator which exhibits high relative reflection contrast (220%) and near zero phase contrast (< π/50) is shown, with potential application to dynamic wavefront control.

Plastic self-adaptation, nonlinear recurrent dynamics and multi-scale memory are desired features in hardware implementations of neural networks, because they enable them to learn, adapt, and process information similarly to the way biological brains do. In this work, these properties occurring in arrays of photonic neurons are experimentally demonstrated. Importantly, this is realized autonomously in an emergent fashion, without the need for an external controller setting weights and without explicit feedback of a global reward signal. Using a hierarchy of such arrays coupled to a backpropagation-free training algorithm based on simple logistic regression, a performance of 98.2% is achieved on the MNIST task, a popular benchmark task looking at classification of written digits. The plastic nodes consist of silicon photonics microring resonators covered by a patch of phase-change material that implements nonvolatile memory. The system is compact, robust, and straightforward to scale up through the use of multiple wavelengths. Moreover, it constitutes a unique platform to test and efficiently implement biologically plausible learning schemes at a high processing speed.

Local and non-local phase-change metasurfaces for dual-function edge-detection/bright-field imaging are designed and simulated. Reconfigurability is via switching phase-change material between crystal and amorphous states. Applications include fast pre-processing for image analysis, optical microscopy and more.

We demonstrate a single step, cost efficient, and residue free UV-interferometric technique for the instantaneous fabrication of high performance, near infrared metagratings with tailored diffraction properties both in terms of dispersion and efficiency.

Active metamaterials are engineered structures that possess novel properties that can be changed after the point of manufacture. Their novel properties arise predominantly from their physical structure, as opposed to their chemical composition and can be changed through means such as direct energy addition into wave paths, or physically changing/morphing the structure in response to both a user or environmental input. Active metamaterials are currently of wide interest to the physics community and encompass a range of sub-domains in applied physics (e.g. photonic, microwave, acoustic, mechanical, etc.). They possess the potential to provide solutions that are more suitable to specific applications, or which allow novel properties to be produced which cannot be achieved with passive metamaterials, such as time-varying or gain enhancement effects. They have the potential to help solve some of the important current and future problems faced by the advancement of modern society, such as achieving net-zero, sustainability, healthcare and equality goals. Despite their huge potential, the added complexity of their design and operation, compared to passive metamaterials creates challenges to the advancement of the field, particularly beyond theoretical and lab-based experiments. This roadmap brings together experts in all types of active metamaterials and across a wide range of areas of applied physics. The objective is to provide an overview of the current state of the art and the associated current/future challenges, with the hope that the required advances identified create a roadmap for the future advancement and application of this field.